Re: [AMBER] RESP charge of the ligand

From: FyD <>
Date: Tue, 16 Oct 2012 10:44:50 +0200

Dear Mu Xia,

> I'm totally confused about the RESP charge for the ligand when
> performing MD calculation. It is recommended that the ligand's
> charge distribution should be obtained by fitting to the
> electrostatic potential calculated by HF/6-31g*. I have several
> questions below.
> 1. Why HF/6-31G* here, why not b3lyp/6-31G*?

This is because HF/6-31G* leads to implicit polarization required in
the additive FF model. See: &

> 2. In ff03, the RESP charges of amino acids are calculated by
> B3LYP/ccpVTZ//HF/6-31G** methods. With ff03 for the protein, should
> we use the same level to obtain the atomic RESP charges of the
> ligand or use HF/6-31G* instead?

Here the strategy followed by Duan et al. is different: the idea is to
use implicit solvation in MEP computation to mimic the environment of
a globular protein.

> 3. If the structure of the ligand is from docking or co-crystal,
> is there any need to optimize it first before calculating the RESP
> charge? If the answer is YES, what is the level? HF/6-31G*,
> B3LYP/6-31G* or the higher the better?

This is highly dependent to the FF you wish to use:
- if you want to use the Cornell et al. FF strategy you should use
HF/6-31G* in MEP computation
- if you want to use the Duan et al. FF strategy you should use
B3LYP/cc-pVTZ SCRF(IEFPCM,Solvent=Ether) as implement in Gaussian 2003.

More generally; See

> 4. When calculating the electrostatic potential, what is the IOp
> settings? I noticed that when one uses antichamber to generate the
> Gaussian input file, the default IOp is "iop(6/33=2, 6/42=6)". But
> someone told me that "iop(6/33=2,6/42=17,6/41=10)" may be better.
> The charges vary a lot when different iop settings are used. What is
> your opinion about this?

The values used by Antechamber are NOT these originally defined in the
default algorithm and these recommended by Kollman & coworkers.
  IOp(6/33=2) only means print the MEP in the Gaussian output
All other IOps should NOT be used if you want to use the default algo.
For instance R.E.D. does _not_ use these IOP [i.e. it only uses
IOp(6/33=2)]; the definition of these IOP can be found from the
Gaussian web site at:

More generally, the answers to your questions can be found in the
article describing the R.E.D. tools:
  & in particular see the "Charge reproducibility" section:

regards, Francois

AMBER mailing list
Received on Tue Oct 16 2012 - 02:00:11 PDT
Custom Search